NORMA eResearch @NCI Library

Festival day arrival prediction using Random Forest, Support Vector Machine and Deep Learning Techniques

Jothimani, Rajbharath (2023) Festival day arrival prediction using Random Forest, Support Vector Machine and Deep Learning Techniques. Masters thesis, Dublin, National College of Ireland.

[thumbnail of Master of Science]
PDF (Master of Science)
Download (4MB) | Preview
[thumbnail of Configuration manual]
PDF (Configuration manual)
Download (1MB) | Preview


Customers of e-commerce business exhibits a varied shopping pattern on any e-commerce portal related to general retail during festival season. Festival sales have begun to make a bigger contribution to the expansion of e-commerce business. With the help of variations in purchase patterns discovered from consumer session information on e-commerce portals, this study aims to predict the arrival of festival using user session data. The main benefit of this study is that it will help e-commerce companies increase their sales on festival days since it can anticipate festival day arrival using the user session details of the e-commerce users. Festival day arrival prediction using machine learning algorithms is made not only to help ecommerce businesses plan their marketing strategies well in advance of the festival, but also serves a number of other purposes, such as restocking specific products with a higher likelihood of selling during that festival and assisting the business in making decisions about improvements to the user interface of the e-commerce portal for obtaining better user traffic. This study will thus empower e-commerce companies to increase their revenue during festivals and initiate action plans towards marketing strategies and product restocking by predicting the arrival of the festival day. Machine learning models such as Random forest classifier, Support vector machine classifier and Deep learning using Keras are used in this research for predicting the festival day arrival since it falls under classification problem. Hyper parameter tuning were performed in these models to optimize the prediction performance. Among the three models chosen, Random forest classifier outperformed both support vector machine classifier and Deep learning using Keras with a higher accuracy of 85.47%. These models are also evaluated based on various other evaluation metrics such as Log loss, Matthew correlation coefficient, precision score, ROC-AUC score, Recall score and F1-score. Random forest classifier not only outperformed the other two models based on accuracy, but also on all the evaluation metrics being mentioned, thus making random forest classifier as the best model for performing the prediction of festival arrival using user session data of e-commerce.

Item Type: Thesis (Masters)
Muntean, Cristina Hava
Uncontrolled Keywords: coefficient; precision; vector; session; recall; strategies; restocking
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science
H Social Sciences > HF Commerce > Electronic Commerce
Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
H Social Sciences > HD Industries. Land use. Labor > Business Logistics > Transportation of Goods and Trade Logistics
Divisions: School of Computing > Master of Science in Data Analytics
Depositing User: Tamara Malone
Date Deposited: 19 May 2023 13:50
Last Modified: 19 May 2023 13:50

Actions (login required)

View Item View Item